Kragujevac Journal of Mathematics

Volume 47(6) (2023), Pages 865–875.

INEQUALITIES FOR MAXIMUM MODULUS OF RATIONAL

FUNCTIONS WITH PRESCRIBED POLES

S. L .WALI1

Abstract. In this paper we prove some results concerning the rational functions
with prescribed poles and restricted zeros. These results in fact generalize or
strengthen some known inequalities for rational functions with prescribed poles
and in turn produce new results besides the refinements of some known polynomial
inequalities. Our method of proof may be useful for proving other inequalities for
polynomials and rational functions.

1. Introduction
Let Pn denote the class of all complex polynomials P (z) := nj=0 cj z j of degree at
P

most n and P 0 (z) be the derivative of P (z). Let Dk− := {z : |z| < k}, Dk+ := {z :
|z| > k} and Tk := {z : |z| = k}. For a function f defined on the circle T1 in the
complex plane C, we write
n
Y
kf k := sup |f (z)|, w(z) := (z − aj )
z∈T1 j=1

and ( )
p(z)
Rn = Rn (a1 , a2 , . . . , an ) := : p ∈ Pn ,
w(z)
where aj ∈ D1+ , j = 1, 2, . . . n.

Key words and phrases. Rational functions, polynomials, Schwarz lemma, inequalities, polar
derivative.
2010 Mathematics Subject Classification. Primary: 26D10. Secondary: 30A10, 30C15.
DOI
Received: March 22, 2020.
Accepted: December 24, 2020.

865
866 S. L. WALI

Thus, Rn is the set of all rational functions with poles a1 , a2 , . . . , an at most and
with finite limit at ∞. We observe that the Blashke product B(z) ∈ Rn , where
n
!
Y 1 − aj z
B(z) := .
j=1 z − aj
For every P ∈ Pn , the following inequality is due to Bernstein [5]:
kP 0 k ≤ nkP k,
where as by an application of maximum modulus principle
kP (R, ·)k ≤ Rn kP k,
where kP (R, ·)k = supz∈TR |P (z)|. Both these inequalities are sharp and equality
holds for polynomials having all zeros at the origin. In case P (z) 6= 0 for z ∈ D1− ,
then we have for z ∈ T1
n
(1.1) kP 0 k ≤ kP k
2
and
Rn + 1
(1.2) kP (R, ·)k ≤ kP k,
2
whereas if P (z) 6= 0 for z ∈ D1+ , then
n
(1.3) kP 0 k ≥ kP k.
2
Inequality (1.1) was conjectured by Erdös and proved by Lax [9], whereas inequality
(1.2) is due to Ankeny and Rivilin [1]. Inequality (1.3) is due to Turán [14]. In all the
inequalities (1.1), (1.2), and (1.3) equality holds for polynomials having all zeros on
the unit disk.
Li, Mohapatra and Rodriguez [10] extended inequalities (1.1) and (1.3) to rational
functions r ∈ Rn and proved the following results.
Theorem 1.1. Suppose r ∈ Rn and all the zeroes of r lie in T1 ∪ D1+ . Then for z ∈ T1
1
|r0 (z)| ≤ |B 0 (z)|krk.
2
Theorem 1.2. Suppose r ∈ Rn , where r has exactly n poles at a1 , a2 , . . . , an and all
the zeros of r lie in T1 ∪ D1− , then for z ∈ T1
1
|r0 (z)| ≥ {|B 0 (z)| − (n − m)}|r(z)|,
2
where m is the number of zeros of r.
The inequality (1.2) was extended to rational functions by Govil and Mohapatra
[7] (see also Aziz and Rather [3]) to read as follows.
Theorem 1.3. Suppose r ∈ Rn and all the zeroes of r lie in T1 ∪ D1+ . Then for z ∈ T1
|B(Rz)| + 1
|r0 (z)| ≤ |r(z)|.
2
 INEQUALITIES FOR MAXIMUM MODULUS OF RATIONAL FUNCTIONS 867

2. Main Results
Theorem 2.1. Suppose r ∈ Rn , where r has n poles at a1 , a2 , . . . , an and all the zeros
of r lie in T1 ∪ D1− with a zero of multiplicity s at origin, then the following inequality
holds for each point z ∈ T1 , such that r(z) 6= 0
zr0 (z)
! ( )
1 |cm | − |cs |
(2.1) Re ≥ |B 0 (z)| + (s + m − n) + ,
r(z) 2 |cm | + |cs |
where m is the number of zeros of r. Inequality (2.1) is sharp and equality holds for
!n
z s (z m−s − 1) 1 − az
r(z) = and B(z) = , z ∈ T1 , a ≥ 1.
(z − a)n z−a
zr0 (z) zr0 (z)
n o
Since r(z)
≥ Re r(z)
, from Theorem 2.1, we immediately have the following.
Corollary 2.1. Suppose r ∈ Rn , where r has n poles at a1 , a2 , . . . , an and all its zeros
lie in T1 ∪ D1− , with s-fold zeros at origin, then for z ∈ T1
( )
0 1 |cm | − |cs |
(2.2) |r (z)| ≥ |B 0 (z)| + (s + m − n) + |r(z)|,
2 |cm | + |cs |
where m is the number of zeros of r. The result is sharp and equality holds for
!n
z s (z m−s − 1) 1 − az
r(z) = and B(z) = ,
(z − a)n z−a
at z = 1 and a ≥ 1.
Note. Inequality (2.2) is trivally true in case r(z) = 0 for z ∈ T1 .
If we take s = 0 in Corollary 2.1, we get the following result, which is an improvement
of Theorem 1.2, earlier proved by Li, Mohapatra and Rodriguez [10, Theorem 4].
Corollary 2.2. Suppose r ∈ Rn , where r has n poles and all the zeros of r lie in
T1 ∪ D1− . Then for z ∈ T1
( )
1 |cm | − |c0 |
|r0 (z)| ≥ |B 0 (z)| − (n − m) + |r(z)|,
2 |cm | + |c0 |
where m is the number of zeros of r. The result is sharp and equality holds for
!n
(z + 1)m 1 − az
r(z) = and B(z) =
(z − a)n z−a
at z = 1 and a ≥ 1.
Since |cm | ≥ |c0 |, therefore as mentioned above, Corollary 2.2 is an improvement of
Theorem 1.2. In case number of poles of r is same as its zeros, that is, when m = n
then Corollary 2.2 gives an improvement of [4, inequality (12)].
Taking ai = α, i = 1, 2, . . . , n, in Corollary 1, we have the following.
868 S. L. WALI

Corollary 2.3. If P (z) is a polynomial of degree m having all zeros in T1 ∪ D1− with
s-fold zero at origin, then
( )
|α| − 1 |cm | − |cs |
(2.3) |Dα P (z)| ≥ m+s+ |P (z)|,
2 |cm | + |cs |
where Dα P (z) := nP (z) + (α − z)P 0 (z) is called the polar derivative of the polynomial
P (z) with respect to the point α and it generalizes the ordinary derivative of P (z) of
degree n in the sense that
Dα P (z)
lim = P 0 (z).
α→∞ α
By taking s = 0 in Corollary 2.3, we get the following sharp result which is also an
extension of a result of Dubinin [6] to the polar derivative of P (z).
Pn j
Corollary 2.4. If P (z) := j=0 cj z is a polynomial of degree n having all zeros in
T1 ∪ D1− , then
( )
|α| − 1 |cn | − |c0 |
(2.4) |Dα P (z)| ≥ n+ |P (z)|.
2 |cn | + |c0 |
Equality in (2.4) holds for a polynomial P (z) = (z−1)n with α > 1. Since |cn | ≥ |c0 |,
it follows that Corollary 2.4 is a refinement of a result of Shah [13].
Dividing both sides of (2.3) by |α| and letting |α| → ∞, we get the following result.
Corollary 2.5. If P ∈ Pn is such that P (z) has all its zeros in T1 ∪ D1− with s-fold
zero at origin, then
!
0 1 |cn | − |c0 |
(2.5) |P (z)| ≥ n+s+ |P (z)|.
2 |cn | + |c0 |
For s = 0 (2.5) reduces to the result of Dubinin [6] and is an improvement of a
classical result of Turán [14].
Next we prove the following refinement of a result of Aziz and Shah [4, Theorem 1].
Theorem 2.2. Suppose r ∈ Rn , where r has exactly n poles a1 , a2 , . . . , an and all the
zeros of r lie in Tk ∪ Dk− , k ≤ 1, with a zero of order s at origin. Then for z ∈ T1
( )
0 1 2(m + sk)
(2.6) |r (z)| ≥ |B 0 (z)| − n + |r(z)|,
2 1+k
where m is the number of zeros of r. The result is sharp and equality holds for
!n
z s (z + k)m−s 1 − az
r(z) = and B(z) =
(z − a)n z−a
at z = 1 and a ≥ 1.
The result of Aziz and Shah [4, Theorem 1] is a special case of Theorem 2.2, if we
take s = 0.
As in previous case, if we take ai = α, i = 1, 2, . . . , n, in Theorem 2.2, we get the
following result on the polar derivatives of a polynomial.
 INEQUALITIES FOR MAXIMUM MODULUS OF RATIONAL FUNCTIONS 869

Corollary 2.6. If P ∈ Pn is such that P (z) 6= 0 in Dk+ , k ≤ 1 with s-fold zero at

origin, then for every α with |α| ≥ 1
n(|α| − 1)(1 + ks)
(2.7) |Dα P (z)| ≥ |P (z)|.
1+k
Remark 2.1. In Corollary 2.6 if we take s = 0, we have the following generalization of
a result of Shah [13].
Corollary 2.7. If P ∈ Pn is such that P (z) has all zeros in Tk ∪ Dk− , then for |α| ≥ 1
and z ∈ T1
n(|α| − 1)
|Dα P (z)| ≥ |P (z)|.
1+k
Remark 2.2. Dividing both sides of (2.7) by |α| and letting |α| → ∞, we have the
following sharp result.
Corollary 2.8. If P ∈ Pn is such that P (z) 6= 0 for z ∈ Dk+ with s-fold zero at origin,
then for z ∈ T1
n + ks
|P 0 (z)| ≥ |P (z)|.
1+k
Equality holds for P (z) = z s (z − k)n−s .
For s = 0, this gives result of Malik [11], whereas for k = 1, s = 0, it reduces to the
classical theorem of Turán [14].
Theorem 2.3. Suppose r ∈ Rn and all the zeros of r lie in Tk ∪ Dk+ . Then for z ∈ T1
(R + k)n
( )
(2.8) |r(Rz)| ≤ |B(Rz)| + 1 krk.
(R + k)n + (1 + Rk)n
Remark 2.3. Theorem 1.3 is a special case of Theorem 2.3, when k = 1.
Remark 2.4. Let w(z) = (z − α)n , |α| > 1, so that
n n
P (z) 1 − αz 1 − αz
Y 
r(z) = and B(z) = = .
(z − α)n 1 z −α z−α
Using this in Theorem 2.3, it can be easily verified that for |α| ≥ R > 1 and z ∈ T1
n
(R + k)n
( )
1 − αRz
|P (Rz)| ≤ + 1 kP k.
(R + k)n + (1 + Rk)n Rz − α
Letting |α| → ∞, we get the following result.
If P (z) is a polynomial of degree n, which does not vanish in |z| < k, k ≥ 1, then
for R > 1 and z ∈ T1
(R + k)n (Rn + 1)
kP (R, ·)k ≤ kP k.
(R + k)n + (1 + Rk)n
This result was earlier proved by Aziz and Mohammad [2].
870 S. L. WALI

3. Lemmas and Proofs

For the proofs of these theorems we need the following lemmas.
Lemma 3.1 ([8]). Let f : D → D be holomorphic. Assume that f (0) = 0. Further
assume that there is a b ∈ ∂D, the boundary of D, so that f extends continuously to
b, |f (b)| = 1 and f 0 (b) exists. Then
2
|f 0 (b)| ≥ .
1 + |f 0 (0)|
The next lemma is due to Aziz and Rather.
Lemma 3.2 ([3]). If r ∈ Rn and z ∈ T1 , then for every R ≥ 0,
|r(Rz)| + |r∗ (Rz)| ≤ {|B(Rz)| + 1}krk,
where r∗ (z) = B(z)r( z1 ).
Proof of Theorem 2.1. Suppose that r(z) 6= 0 for z ∈ T1 and all the poles of r(z) lie
in D1+ . Since r(z) has a zero at origin of multiplicity s. Therefore,
P (z) z s h(z)
r(z) = = ,
w(z) w(z)
where
m−s m−s
cs+j z j = cm zj ∈ D1− , j = 1, 2, . . . , m − s,
X Y
h(z) := (z − zj ),
j=0 j=1

and
n
Y
w(z) = (z − aj ).
j=1
This gives
r0 (z) s h0 (z) w0 (z)
= + − .
r(z) z h(z) w(z)
Equivalently,
zr0 (z) zh0 (z) zw0 (z)
! ! !
(3.1) Re = s + Re − Re .
r(z) h(z) w(z)
Since h(z) has all zeros in D1− , therefore
1
 
∗ m−s
h (z) = z h
z
has all zeros in D1+ , and hence
cm m−s
!
zh(z) Y z − zj
(3.2) G(z) = ∗ =z
h (z) cm j=1 1 − zzj
is analytic in T1 ∪ D1− , with G(0) = 0 and |G(z)| = 1 for z ∈ T1 .
 INEQUALITIES FOR MAXIMUM MODULUS OF RATIONAL FUNCTIONS 871

Applying Lemma 3.1 to G(z), we get for z ∈ T1

2
(3.3) |G0 (z)| ≥ .
1 + |G0 (0)|
Since
m−s
0
Y |cs |
|G (0)| = zj = ,
j=1 |cm |
it can be easily verified (see [15, proof of Lemma 1]) that for every z ∈ T1
zh0 (z)
( )
m−s−1 |cm |
(3.4) Re ≥ + .
h(z) 2 |cm | + |cs |
Again we have
w∗ (z)
B(z) = ,
w(z)
where !
∗ 1 n
w (z) = z w .
z
This gives (see [15, Lemma 1])
zw0 (z) n − |B 0 (z)|
( )
(3.5) Re = .
w(z) 2
Now using (3.4) and (3.5) in (3.1), we conclude that
zr0 (z)
( ) ( )
1 |cm | − |cs |
Re ≥ s + |B 0 (z)| − (n − m) + .
r(z) 2 |cm | + |cs |
The proof of Theorem 2.1 is completed. 
Proof of Theorem 2.2. Suppose that for each point z ∈ T1 , r(z) 6= 0 and all the poles
of r(z) lie in D1+ . Since r(z) has a zero of order s at origin, therefore
P (z) z s Q(z)
r(z) = = ,
w(z) w(z)
where
m−s
cs+j z j
X
Q(z) =
j=0
is a polynomial of degree m − s having all zeros in |z| ≤ k. This gives
zr0 (z) zQ0 (z) zw0 (z)
(3.6) Re = s + Re − Re .
r(z) Q(z) w(z)
m−s
(z − zj ), |zj | ≤ k ≤ 1, and it can be easily verified that
Q
We write Q(z) = cm
j=1

zQ0 (z) m−s

Re ≥ .
Q(z) 1+k
872 S. L. WALI

Also, as in (3.5)
zw0 (z) n − |B 0 (z)|
( )
Re = .
w(z) 2
Using this in (3.6), we get
zr0 (z) m − s n − |B 0 (z)|
Re ≥s+ − .
r(z) 1+k 2
Now for z ∈ T1 such that r(z) 6= 0, we have
zr0 (z) zr0 (z)
( )
1 2(m + sk) − n(1 + k)
≥ Re ≥ |B 0 (z)| + .
r(z) r(z) 2 1+k
This gives for z ∈ T1 such that r(z) 6= 0,
( )
0 1 2(m + sk)
|r (z)| ≥ |B 0 (z)| − n + |r(z)|.
2 1+k
Since the result is trivially true if r(z) = 0 for z ∈ T1 , it follows that
( )
0 1 2(m + sk)
|r (z)| ≥ |B 0 (z)| − n + |r(z)|,
2 (1 + k)
for all z ∈ T1 . The proof of Theorem 2.2 is completed. 
Proof of Theorem 2.3. Since zeros of r(z) lie in Tk ∪ Dk+ , therefore all zeros of P (z)
lie in Tk ∪ Dk+ , k ≥ 1 and w(z) = nj=1 (z − αj ), |αj | > 1 for all j = 1, 2, . . . , n.
Q

Also
n
P ( z1 )
! !
1 1 − αj z
 
∗
Y
r (z) = B(z)r =
z j=1 z − αj w( z1 )
n
z n P ( z1 ) z n P ( z1 ) P ∗ (z)
!
Y 1 − αj z
= n = n = .
j=1 z − αj Q
(1 − αj z)
Q
(z − αj ) w(z)
j=1 j=1

Therefore,
r(z) P (z)
(3.7) ∗
= ∗ .
r (z) P (z)
We write
n
(z − rj eiθj ),
Y
P (z) = where rj ≥ k ≥ 1, j = 1, 2, . . . , n,
j=1
n
so that P ∗ (z) = z n P ( z1 ) = (1 − zrj eiθj ).
Q
j=1
For 0 ≤ θ < 2π and R > 1 we have
n
P (Reiθ ) Y Reiθ − rj eiθj
(3.8) = .
P ∗ (Reiθ ) j=1 1 − rj Re
i(θ−θj )
 INEQUALITIES FOR MAXIMUM MODULUS OF RATIONAL FUNCTIONS 873

Now
n 2 n 2
Y Reiθ − rj eiθj Y Rei(θ−θj ) − rj
=
j=1 1 − rj Rei(θ−θj ) j=1 1 − rj Rei(θ−θj )
n
Rei(θ−θj ) − rj Re−i(θ−θj ) − rj
!
Y
= ·
j=1 1 − rj Rei(θ−θj ) 1 − rj Re−i(θ−θj )
!2
n
Y R2 − 2Rrj cos(θ − θj ) + rj2 Yn
R + rj
= 2 2
≤ .
j=1 1 − 2Rrj cos(θ − θj ) + rj R j=1 1 + Rrj

Therefore, from (3.8), we have

2 n
!2
P (Reiθ ) Y R + rj
(3.9) ≤ .
P ∗ (Reiθ ) j=1 1 + Rrj
Since |rj | ≥ k, k ≥ 1, it can be easily verified that
R + rj R+k
≤ .
1 + Rrj 1 + Rk
Using this in (3.9), we get
n
!n
P (Reiθ )
!
Y R+k R+k
(3.10) ∗ iθ
≤ = .
P (Re ) j=1 1 + Rk 1 + Rk
Combining (3.7) and (3.10), we get
!n
r(Rz) R+k
≤ , for z ∈ T1 , k ≥ 1, R > 1.
r∗ (Rz) 1 + Rk
Equivalently,
!−n
R+k
(3.11) |r(Rz)| ≤ |r∗ (Rz)|.
1 + Rk
Now using Lemma 3.2, we get from (3.11)
( !−n )
R+k
+ 1 |r(Rz)| ≤ |r(Rz)| + |r∗ (Rz)|.
1 + Rk
That is,
(R + k)n
|r(Rz)| ≤ {|B(Rz)| + 1}|r|.
(R + k)n + (1 + Rk)n
The proof of Theorem 2.3 is completed. 
Proof of Corollary 2.3. Since r(z) has a pole of order n at z = α, |α| > 1, we have
P (z)
r(z) = .
(z − α)n
874 S. L. WALI

Therefore,
(z − α)n P 0 (z) − n(z − α)n−1 P (z) −[nP (z) + (α − z)P 0 (z)] −Dα P (z)
r0 (z) = = n+1
= .
(z − α)2n (z − α) (z − α)n+1
Also, !n
1 − αz
B(z) =
z−α
gives
n(1 − αz)n−1 (|α|2 − 1)
B 0 (z) = .
(z − α)n+1
Using these facts, we immediately get for z ∈ T1 from inequality (2.2)
1 n|z − α|n−1 (|α|2 − 1)
( !)
Dα P (z) |cm | − |cs | |P (z)|
n+1
≥ n+1
+ s+m−n+ .
(z − α) 2 |z − α| |cm | + |cs | |z − α|n
This gives
( ! )
1 n(|α|2 − 1) |cm | − |cs |
|Dα P (z)| ≥ + s+m−n+ |z − α| |P (z)|
2 |z − α| |cm | + |cs |
( !)
(|α| − 1) |cm | − |cs |
≥ n+ s+m−n+ |P (z)|
2 |cm | + |cs |
( )
(|α| − 1) |cm | − |cs |
= s+m+ |P (z)|, z ∈ T1 , |α| ≥ 1.
2 |cm | + |cs |
Using the argument of continuity in case of poles the proof of Corollary 2.3 completes.

Proof of Corollary 2.6. Since we have
!n
P (z) 1 − αz
r(z) = and B(z) = ,
(z − α)n z−α
therefore
−Dα P (z)
r0 (z) =
(z − α)n+1
and
n(1 − αz)n−1 (|α|2 − 1)
B 0 (z) = .
(z − α)n+1
Using in inequality (2.6), with m = n, we get
1 n|z − α|n−1 (|α|2 − 1) n(1 − k) + 2ks |P (z)|
( )
Dα P (z)
≥ + .
(z − α)n+1 2 |z − α|n+1 1+k |z − α|n
This gives
( )
1 n(|α|2 − 1) n(1 − k) + 2ks
|Dα P (z)| ≥ + |z − α| |P (z)|
2 |z − α| 1+k
 INEQUALITIES FOR MAXIMUM MODULUS OF RATIONAL FUNCTIONS 875

( )
1 n(1 − k) + 2ks
≥ n(|α| − 1) + (|α| − 1) |P (z)|
2 1+k
n(|α| − 1)(1 + ks)
= |P (z)|, z ∈ T1 , |α| ≥ 1.
1+k
Using the argument of continuity in case of poles the proof of Corollary 2.6 completes.


References
[1] N. C. Ankeny and T. J. Rivlin, On a theorem of S. Bernstein, Pacific J. Math. 5(2) (1955),
849–852.
[2] A. Aziz and Q. G. Mohammad, Growth of polynomials, with zeros outside a circle, Proc. Amer.
Math. Soc. 81(4) (1981), 549–553. https://doi.org/10.2307/2044157
[3] A. Aziz and N. A. Rather, Growth of maximum modulus of rational functions with prescribed
poles, Math. Inequal. Appl. 2 (1999), 165–173.
[4] A. Aziz and W. M. Shah, Some properties of rational functions with prescribed poles and restricted
zeros, Math. Balkanica (N.S) 18 (2004), 33–40.
[5] S. Bernstein, Sur la limitation des dérivées des polynomes, C. R. Math. Acad. Sci. Paris 190
(1930), 338–340.
[6] V. N. Dubinin, Applications of the Schwarz lemma to inequalities for entire functions with
constraints on zeros, J. Math. Sci. (N.Y) 143(3) (2007), 3069–3076. https://doi.org/10.1007/
s10958-007-0192-4
[7] N. K. Govil and R. N. Mohapatra, Inequalities for Maximum Modulus of Rational Functions with
Prescribed Poles, Dekker, New York, 1998.
[8] S. G. Krantz, The Schwarz bemma at the boundary, Complex Var. Elliptic Equ. 56(5) (2011),
455–468. https://doi.org/10.1080/17476931003728438
[9] P. D. Lax, Proof of a conjecture of P. Erdös on the derivative of a polynomial, Bull. Amer. Math.
Soc. 50 (1944), 509–513. https://doi.org/10.1090/s0002-9904-1944-08177-9
[10] X. Li, R. N. Mohapatra and R. S. Rodriguez, Bernstien-type inequalities for rational functions
with prescribed poles, J. London. Math. Soc. 1 (1995), 523–531. https://doi.org/10.1112/
jlms/51.3.523
[11] M. A. Malik, On the derivative of a polynomial, J. London. Math. Soc. 1 (1969), 57–60. https:
//doi.org/10.1112/jlms/s2-1.1.57
[12] R. Osserman, A sharp Schwarz inequality on the boundary for functions regular in the disk,
Proc. Amer. Math. Soc. 12 (2000), 3513–3517. https://doi.org/10.1090/s0002-9939-1988.
0928994-0
[13] W. M. Shah, A generalisation of a theorem of Paul Túran, J. Ramanujan Math. Soc. 11(1)
(1996), 67–72.
[14] P. Turán, Über die ableitung von polynomen, Compos. Math. 7 (1939), 89–95.
[15] S. L. Wali and W. M. Shah, Applications of the Schwarz lemma to inequalities for ra-
tional functions with prescribed poles. J. Anal. 25 (2017), 43–53. https://doi:10.1007/
s41478-016-0025-2
1
Department of Mathematics,
Central University of Kashmir,
Ganderbal Jammu and Kashmir, India
Email address: shahlw@yahoo.co.in